The Structure of the Hexameric Atrazine Chlorohydrolase Atza

research papers

The structure of the hexameric atrazine chlorohydrolase AtzA

ISSN 1399-0047 T. S. Peat,a J. Newman,a S. Balotra,b,c D. Lucent,d A. C. Wardenc and C. Scotta*

aCSIRO Biomedical Manufacturing, Parkville, Australia, bResearch School of Chemistry, Australian National University, Canberra, Australia, cCSIRO Land and Water Flagship, Black Mountain, Canberra, Australia, and dDivision of Engineering and Physics, Wilkes University, Wilkes-Barr, Pennsylvania, USA. *Correspondence e-mail: [email protected] Received 14 November 2014 Accepted 12 January 2015 Atrazine chlorohydrolase (AtzA) was discovered and purified in the early 1990s from soil that had been exposed to the widely used herbicide atrazine. It was subsequently found that this enzyme catalyzes the first and necessary step in Keywords: atrazine chlorohydrolase. the breakdown of atrazine by the soil organism Pseudomonas sp. strain ADP. Although it has taken 20 years, a crystal structure of the full hexameric form of PDB references: AtzA, 4v1x; 4v1y AtzA has now been obtained. AtzA is less well adapted to its physiological role (i.e. atrazine dechlorination) than the alternative metal-dependent atrazine Supporting information: this article has supporting information at journals.iucr.org/d chlorohydrolase (TrzN), with a substrate-binding pocket that is under considerable strain and for which the substrate is a poor fit.

1. Introduction The last century saw the introduction of a host of anthropogenic compounds, such as pesticides and antibiotics, into the environment (Russell et al., 2011; Copley, 2000). This inﬂux of new compounds has introduced a range of new selection pressures through factors such as toxicity and the availability of abundant potential nutrient sources (Copley, 2000; Russell et al., 2011; Wackett, 2009). Nature has evolved mechanisms to cope with these new selection pressures, including the estab- lishment of new metabolic pathways and enzymes. These new metabolic pathways and their associated enzymes are a tremendous resource for advancing our understanding of the mechanisms and constraints that underpin the evolutionary process, particularly with respect to the acquisition of new enzymatic functions. The bacterial atrazine catabolism pathway from Pseudo- monas sp. strain ADP is a particularly well studied example of a metabolic pathway that has evolved in response to human perturbations of the chemical composition of the environment. The pathway is comprised of six enzymatic steps (Fig. 1): the sequential hydrolysis of the chloride and two N-alkyl chains to produce cyanuric acid by AtzA (de Souza et al., 1996; Scott et al., 2009), AtzB (Boundy-Mills et al., 1997; Seffernick et al., 2007) and AtzC (Sadowsky et al., 1998; Shapir et al., 2002), followed by a ring-opening hydrolysis (AtzD; Fruchey et al., 2003; Seffernick et al., 2012; Peat et al., 2013) and two deamination steps (AtzE and AtzF; Balotra et al., 2014, 2015; Shapir, Sadowsky et al., 2005; Cameron et al., 2011; Martinez et al., 2001). The catabolism of cyanuric acid possibly predates the introduction of anthropogenic herbicides, as cyanuric acid is a naturally occurring compound (albeit not an abundant one; Wackett, 2009). However, the enzymes that catalyze the ﬁrst three steps of the pathway (AtzA, AtzB and AtzC) are likely to have evolved recently in response to the presence of atrazine in the environment (Wackett, 2009; Udikovic´-Kolic´ et al., 2013).

710 doi:10.1107/S1399004715000619 Acta Cryst. (2015). D71, 710–720 research papers

The atrazine chlorohydrolase AtzA has attracted a great There have been studies that have used DNA shufﬂing deal of attention as a model for studying the evolutionary (Raillard et al., 2001) and evolutionary trajectory reconstruc- process, not least because of its relationship to melamine tion (Noor et al., 2012) to understand how these large func- deaminase (TriA). AtzA and TriA share 98% identity, tional differences could have evolved with so few genetic differing by just nine amino acids over their 470-amino-acid changes. length, and each difference is conferred by one of just nine In other bacterial genera, particularly Gram-positive point mutations (Seffernick et al., 2001; Noor et al., 2012). bacteria (e.g. Arthrobacter and Nocardioides), the role of However, AtzA and TriA have quite distinct enzymatic AtzA is occupied by the alternative chlorohydrolase TrzN functions, with AtzA only capable of hydrolytic dehalogena- (Mulbry et al., 2002; Sajjaphan et al., 2004; Shapir et al., 2006; tion (Fig. 1) and the deaminase TriA possessing only low levels Seffernick et al., 2010). Both AtzA and TrzN belong to the of promiscuous dehalogenase activity (Seffernick et al., 2001). same large family of amidohydrolases, although they are so different physically and phylogenetically that it is likely that atrazine chlorohydrolase activity evolved independently in each enzyme. While AtzA is a hexamer that contains one essential Fe2+ per monomer (de Souza et al., 1996; Scott et al., 2009), TrzN (which is 25% identical to AtzA) is a dimer that has a single Zn2+ bound in each active site (Shapir et al., 2006). Moreover, while AtzA can only hydrolyze triazine halides, TrzN can hydrolyze a broad range of substituents (including

halide, –OCH3 and –SCH3 groups) from both triazines and pyrimidines (Shapir, Rosendahl et al., 2005; de Souza et al., 1996). TrzN is also an order of magnitude more efﬁcient than 5 1 1 AtzA, with a kcat/Km value of 10 s M compared with 1.5 104 s1 M1. This difference is in large part owing to

the relatively high Km of AtzA, which is greater than the water solubility of atrazine (153 mM; Scott et al., 2009), compared with that of TrzN for atrazine (20 mM; Jackson et al., 2014). In contrast to most other known hydrolytic dehalogenases, which use an active-site carboxylic acid (Asp) to displace the halide ion (Verschueren et al., 1993; Newman et al., 1999), the metal-dependent reaction mechanisms of AtzA and TrzN make these two enzyme lineages somewhat unusual in nature. Despite intensive genetic and biochemical study of AtzA, the lack of an experimentally derived structural model has hampered efforts to understand the details of the sequence– function relationship. Here, we present the ﬁrst X-ray structure of AtzA, upon which we have based a reinterpretation of the genetic and biochemical analyses of this model chlorohydrolase.

2. Materials and methods 2.1. Protein expression and purification The construction of the pCS150 expression vector containing the wild-type atzA gene has been described elsewhere (Scott et al., 2009). The mutant atzA gene encoding the AtzA Ala170Thr, Met256Ile, Pro258Thr, Tyr261Ser variant was obtained from Life Technologies (Australia) and provided in pMK-RQ with NdeI and BamHI sites placed for subcloning into the pCS150 vector. pCS150 plasmids containing either the wild-type or mutant atzA genes were used to transform elec- trocompetent Escherichia coli BL21 DE3 cells (Invitrogen) and were grown at 310 K on Luria–Bertani (LB; Lennox, 1955) agar [1.5%(w/v)] plates supplemented with 100 mgml1 Figure 1 ampicillin. Overnight starter cultures of 50 ml were inoculated Atrazine catabolic pathway. The six hydrolytic steps of the atrazine catabolic pathway of Pseudomonas sp. strain ADP as catalysed by AtzA, with a single colony. The overnight cultures were diluted 1:20 AtzB, AtzC, AtzD, AtzE and AtzF are shown. into 950 ml LB medium and were shaken at 310 K and

Acta Cryst. (2015). D71, 710–720 Peat et al. AtzA 711 research papers

1 200 rev min until an OD600 of 0.6–0.8 was obtained. Protein Table 1 expression was initiated by the addition of 100 mM isopropyl Data-collection and refinement statistics. -d-1-thiogalactopyranoside (IPTG). The induced cultures Values in parentheses are for the highest resolution shell. 1 were kept at 310 K overnight whilst shaking at 200 rev min . PDB code 4v1y 4v1x The cells were then harvested by centrifugation at 4000g for 10 min in an Avanti J-E centrifuge (Beckman Coulter, India- Space group P22121 P212121 Unit-cell parameters (A˚ , ) a = 117.5, b = 195.6, a = 117.3, b = 146.1, napolis, USA), resuspended in lysis buffer (50 mM HEPES, c = 283.9, c = 196.4, pH 7.5, 100 mM NaCl) and lysed by passage through an = = = 90.0 = = = 90.0 ˚ Avestin C3 homogenizer three times at 124 MPa. Insoluble Resolution (A) 49.5–2.80 (2.85–2.80) 48.7–2.20 (2.24–2.20) Rmerge 0.229 (0.906) 0.235 (1.151) cellular debris was removed by centrifugation at 21 000g using Rmeas 0.266 (1.042) 0.249 (1.223) an Avanti J-E centrifuge. Rp.i.m. 0.135 (0.531) 0.090 (0.438) AtzA and the AtzA variant were purified from lysed cells in CC1/2 0.989 (0.795) 0.995 (0.840) hI/(I)i 8.8 (2.6) 11.6 (3.2) three steps: His-tag affinity chromatography using an Ni–NTA Completeness (%) 100 (100) 99.6 (99.2) Superflow Cartridge (Qiagen, Maryland, USA) with a Multiplicity 7.5 (7.6) 15.1 (15.3) gradient of 0–300 mM imidazole in 50 mM HEPES pH 7.5, Refinment Resolution (A˚ ) 49.5–2.8 48.7–2.2 100 mM NaCl and size-exclusion chromatography using a Unique reflections 153180 170140 130 ml column packed with Superdex 200 prep-grade resin Rwork/Rfree (%) 19.3/22.2 16.6/19.6 (GE Healthcare Life Sciences, Australia) with a buffer No. of atoms Total 44810 23156 composed of 50 mM HEPES pH 7.5, 100 mM NaCl. The His Protein 44565 22221 tag was removed by thrombin proteolysis and the protein Other 105 42 was then purified by size-exclusion chromatography with a Ions 15 6 Water 126 887 Superdex 200 column. The protein was concentrated in an B factors (A˚ 2) Amicon Ultra-15 Centrifugal Filter Unit with an Ultracel-30 Overall 29.2 20.1 membrane (Millipore, Carrigtwohill, Ireland) to 11.6 mg ml1 Protein 29.4 19.7 Other 32.4 36.7 and snap-frozen in liquid nitrogen in 100 ml aliquots. The final Ions 29.2 24.3 purity was estimated to be 98% from a Coomassie-stained gel Water 11.8 22.8 and typical yields were 5–7 mg purified protein from 1 l of LB R.m.s. deviations Bond lengths (A˚ ) 0.014 0.017 medium. Bond angles () 1.613 1.675

was then rebuilt by hand using Coot (Emsley et al., 2010) and 2.2. Crystallization and structure solution reﬁned using REFMAC (Murshudov et al., 2011). Two full All crystallization experiments were performed in 96-well hexamers were found to be present in the asymmetric unit,

SD-2 plates (IDEX, USA) against a 50 ml reservoir. Initial each being a trimer of dimers. The model refined to Rwork and crystals were obtained using droplets consisting of 150 nl Rfree values of 19.3 and 22.2%, respectively (Table 1). The concentrated protein solution (11 mg ml1) combined with Ramachandran plot (from Coot) shows 94.9% of the residues 150 nl reservoir solution and these were used in seeding in the most favourable region, 3.9% in the allowed region and experiments. The final crystals were obtained using droplets 1.1% in the outlier region. After this original structure was consisting of 150 nl concentrated protein solution combined built, a second set of crystals were obtained under similar with 120 nl reservoir solution and 30 nl seed stock. Either a conditions [50 mM HEPES pH 7.3, 4.6%(w/v) PEG 10 000] Phoenix (ARI, USA) or a Mosquito (TTP Labtech, UK) robot and another data set was obtained on the MX-2 beamline with was used to place the crystallization drops. Crystals (Supple- data that extended to 2.2 A˚ resolution. This crystal was mentary Fig. S1) grew unreliably from a reservoir consisting of determined to belong to space group P212121 and a single 5.5%(w/v) PEG 8000, 2.7%(v/v) diethylene glycol, 50 mM hexamer was found in the asymmetric unit. The model and HEPES pH 7.1 at 281 K and were used to collect X-ray data structure factors for both structures have been deposited in on the MX-2 beamline of the Australian Synchrotron at a the PDB as entries 4v1x (2.2 A˚ resolution) and 4v1y (2.8 A˚ wavelength of 0.9537 A˚ (13 000 eV). The crystals were cryo- resolution). protected by the addition of reservoir solution supplemented with a further 20% diethylene glycol prior to cryocooling in 2.3. Molecular modelling liquid N2. Data were indexed with XDS (Kabsch, 2010) and scaled using AIMLESS (Evans, 2011). The structure was The gas-phase equilibrium conformation of the AtzA solved using molecular replacement (Phaser; McCoy et al., metal-coordination site was calculated using density func- 2007) with PDB entry 3hpa (32% sequence identity, 15 gaps; tional theory with the B3LYP functional and the 6-31+G* Hall et al., 2010). A clear solution was only found when a basis set (with the LANL2DZ pseudopotential used for iron dimer of 3hpa was used, with six dimers placed in the core electrons) as implemented in Gaussian 09 (Frisch et al., asymmetric unit. The space group was found to be P22121 and 2009). the resolution of the data extended to 2.8 A˚ (see Table 1). The Atrazine was docked into a flexible AtzA binding pocket model was initially built using Buccaneer (Cowtan, 2006) and using the RosettaLigand docking protocol with the coordi-

712 Peat et al. AtzA Acta Cryst. (2015). D71, 710–720 research papers nates from PDB entry 4v1x. Partial atomic charges were (according to protein–ligand interaction energy) were calculated for the ligands using the AM1BCC Hamiltonian as considered for final examination. implemented in QuacPac (v.1.6.3.1; OpenEye Scientific Soft- ware, Santa Fe, New Mexico, USA) and conformers were 2.4. Dynamic light scattering enumerated using Omega (v.2.5.1.4; OpenEye Scientific Soft- Concentrated AtzA (11.6 mg ml1) was diluted by 50% in ware, Santa Fe, New Mexico, USA). 10 000 docking trajec- a serial fashion with 50 mM HEPES pH 7.5, 100 mM NaCl to tories were performed using the RosettaLigand docking obtain a range of concentrations from 11.6 to 0.09 mg ml1. protocol (Davis et al., 2009). The output ensembles were culled 20 ml volumes of each sample, in duplicate, were placed in a to retain the top 5% of structures according to the total energy 384-well plate with blanks consisting of just buffer alone. The (computed from the Rosetta energy function). From this plate was placed in a DynaPro plate reader DLS machine smaller ensemble, the top 20 protein–ligand complexes (Wyatt Technology, Santa Barbara, California, USA) and 50

Figure 2 Structure of AtzA. (a) The X-ray structure of the AtzA hexamer is a trimer of dimers and is coloured by monomer. The structures of the dimer (b) (coloured by monomer), monomer (c) (coloured by secondary structure) and active site (d) are shown. The active-site metal (Fe2+) is shown as an orange sphere.

Acta Cryst. (2015). D71, 710–720 Peat et al. AtzA 713 research papers spectra were obtained at 297 K for each sample and averaged. The AtzA hexamer (about 315 kDa in molecular weight) The top three concentrations gave a radius of gyration of is a trimer of dimers (Fig. 2a), with the dimer interface being 5.9 nm, whereas the lowest concentration (0.09 mg ml1) gave significantly larger than the individual protomer interfaces an average radius of 5.2–5.3 nm. Measuring the hexamer used to make up the hexamer (Figs. 2a and 2b): 3295 A˚ 2 always gave a diameter of greater than 10 nm (10.5 to compared with 585 A˚ 2. The hexamer was confirmed by DLS 13.0 mm, depending on the points chosen), whereas the and size-exclusion chromatography (SEC; Supplementary Fig. longest dimension seen for a dimer was 8.3 nm. S2), consistent with previous reports for the solution state of the protein (de Souza et al., 1996; Scott et al., 2009). The dimer 3. Results and discussion surface covers 17–18% of the monomer (a single monomer has a surface area of 18 600 A˚ 2;Fig.2c). The dimer interface is 3.1. The structure of hexameric AtzA similar to that found in the other amidohydrolases and the Analysis with PDBeFold revealed that the closest structure dimer is similar enough in overall structure that the search in the PDB to AtzA was that used for molecular replacement, model used for molecular replacement (PDB entry 3hpa) gave PDB entry 3hpa (r.m.s.d. of 1.6 A˚ ;Hallet al., 2010), which was a significantly more robust solution when the dimer was used also the structure with the greatest amino-acid sequence instead of the monomer. Two full hexamers (over 600 kDa) identity (32% over 411 residues). Another structural genomics were found in the asymmetric unit of the crystal structure in target, PDB entry 3lnp (Kube et al., 2013), is the next closest space group P22121, whereas a single hexamer was found in by structural homology, followed by two adenosine deami- space group P212121. The two structures are essentially iden- nases from Xanthomonas campestris (PDB entry 4dzh; New tical (r.m.s.d. of 0.3 A˚ ), with the higher resolution data giving York Structural Genomics Research Consortium, unpublished greater clarity in amino-acid positions and having significantly work) and Pseudomonas aeruginosa (PDB entry 3pao; New more water molecules added to the model. The most obvious York Structural Genomics Research Consortium, unpublished interaction seen when looking down the threefold axis of the work). The TrzN structure (Seffernick et al., 2010; PDB entry hexamer is the 11-amino-acid insertion (relative to PDB entry 3lsb), which has 26% sequence identity over 414 residues and 3hpa) that forms the loop and helix encompassing residues an r.m.s.d. of 1.8 A˚ , is the fifth most similar structure to AtzA 160–183 that has not been seen in the other amidohydro- in the PDB. All of these molecules are classified as amido- lase structures to date. The ‘cavity’ in the hexamer is hydrolases, with 3hpa proposed to be an 8-oxoguanine approximately 14 A˚ across at the entrance, with one residue in deaminase. Although 3lnp has a highly homologous active particular, Arg163, forming much of the surface of this site (four histidine residues and an aspartic acid residue), a entrance (Fig. 3). calcium ion was modelled into the density instead of either The atzA genes from Aminobacter aminovorans isolates Fe2+ or Zn2+; all of the other top hits on the list have either recently collected from vineyards in France (Noor et al., 2014) Fe2+ or (more commonly) Zn2+ in their active sites. have accumulated a variety of mutations that have been

Figure 3 The entrance to the central cavity of the AtzA hexamer and the amino-acid residues of the hexamer-stabilizing interface. The entrance to the AtzA hexamer central cavity is largely formed by Arg163 contributed by each monomer. The hole formed in the hexamer is approximately 14 A˚ across at the entrance. Amino-acid substitutions at residues Ala170, Met256, Pro258 and Tyr261 destabilize the hexamer. The interactions between three monomers on one face of the hexamer (left) and at a single monomer–monomer interface (right) are shown.

714 Peat et al. AtzA Acta Cryst. (2015). D71, 710–720 research papers

demonstrated to confer changes in substrate speciﬁcity and alter 2+ the Kd of the enzyme for Fe . Interestingly, four of the amino- acid changes conferred by these mutations (Ala170Thr, Met256Ile, Pro258Thr and Tyr261Ser) are located in the interface responsible for coordinating the AtzA hexamer (Fig. 3). AtzA variants carrying all four amino-acid substitutions were observed in the A. aminovorans isolates in the study by Noor et al. (2014). As these amino acids could effect a change in hexamer formation, the AtzA Ala170Thr, Met256Ile, Pro258Thr, Tyr261Ser variant was expressed and puri- ﬁed and its size was determined by SEC. Unlike the wild-type Figure 4 Access to the active site of AtzA. The surface of the active site and solvent-access channel (which form a protein, which eluted from SEC single continuous channel from the active-site metal to the solvent) is shown in an AtzA monomer which with an estimated molecular has been coloured by secondary structure. The amino acids that form this surface (His66, His68, Gln71, weight of 300 kDa, the variant Phe84, Tyr85, Trp87, Leu88, Phe89, Val92, Tyr93, Asp128, Met155, Phe157, Met160, Asp161, Ile164, eluted with an estimated mole- Gln165, Val168, Leu180, Ser182, Ile183, Met184, Ala216, Thr217, Thr219, Ala220, His243, Glu246, Asp250, His276, Leu305, Asp327, Asn328 and Ser331) are shown as sticks, but have not been labelled for clarity. The cular weight of 100 kDa (i.e. a amino-acid resides are labelled in Supplementary Fig. S5. The active-site metal (Fe2+) is shown as an orange dimer; data not shown). This sphere. The images are rotated 180 around the vertical axis with respect to each other. suggests that there is some factor that has selected dimeric AtzA in the more recent French isolates when compared with the AtzA from the original Pseudo- monas isolate obtained from the USA, although it is uncertain whether the dimeric AtzA variants predate the hexameric form or vice versa.

3.2. Analysis of the AtzA active site The substrate-binding pocket and catalytic centre of each monomer is accessible via a long hydrophobic channel. The channel, substrate-binding pocket and active site are comprised of His66, His68, Gln71, Phe84, Tyr85, Trp87, Leu88, Phe89, Val92, Tyr93, Asp128, Met155, Phe157, Met160, Asp161, Ile164, Gln165, Val168, Leu180, Ser182, Ile183, Met184, Ala216, Thr217, Thr219, Ala220, His243, Glu246, Asp250, His276, Leu305, Asp327, Asn328 and Ser331 (Fig. 4). The identities of these residues are in good agreement with a homology model used to successfully predict sites for muta-

genesis to substantially reduce the Km for atrazine (i.e. Ala216, Thr217, Thr218, Ala219 and Asp250; Scott et al., 2009). The mononuclear active-site metal is coordinated by His66, His68, His243, His276 and Asp327 (Fig. 2d); it is a subtype III Figure 5 amidohydrolase metal-binding motif similar to that found in Density found in the active site of AtzA. A composite OMIT map was E. coli adenine deaminase (Seibert & Raushel, 2005). The generated using the CCP4 package and the ﬁgure shows the ‘extra’ bond lengths between the bound cation and ligating amino- density seen in the active site which may be owing to atrazine (top of the acid side chains are longer than would be expected for a metal ﬁgure, blue mesh). The active-site iron is shown as an orange sphere, water is shown as a red sphere and several active-site residues are shown centre of this type, with lengths in the following ranges: Fe– with associated density. The OMIT map was set to 1. His66, 2.68–2.88 A˚ ; Fe–His68, 2.76–2.96 A˚ ; Fe–His243, 3.01–

Acta Cryst. (2015). D71, 710–720 Peat et al. AtzA 715 research papers

3.14 A˚ ; Fe–His276, 2.76–2.99 A˚ ; Fe–Asp327, 2.41–2.49 A˚ . may be mobile in order to allow substrate entry or exit. One Although the conformations of the metal-coordinating side other Ramachandran outlier, Thr368, is interesting as it is in chains found in the active site are indicative of an octahedral the middle of a helix; the preceding residue, Ala367, which is complex (as would be expected for Fe2+ binding), the also part of the helical structure, has its carbonyl hydrogen- geometry is distorted from that predicted by gas-phase elec- bonded to another residue (Ile63) instead of being in line with tronic structure calculations (Supplementary Fig. S3). Iron- the other carbonyls in the helix. coordinating histidine residues at these distances are observed Examination of the binding pocket indicated that it might in other structures in the PDB, although they are far less be difﬁcult for atrazine to bind in an orientation that would frequent than those with shorter bond lengths (the mean His– permit nucleophilic substitution by an activated water (as is Fe bond length is 2.2 A˚ , with a standard deviation of 0.2 A˚ ; the case in TrzN). Atrazine was both soaked into crystals and Supplementary Fig. S4). Although no negative density was co-crystallized with the protein, with the co-crystallization seen when the Fe atoms were set to 100% occupancy, the experiment resulting in crystals that diffracted to 2.2 A˚ reso- resulting B factors were over double those for the nearby lution. However, only a small amount of broken density was residues. As a result, we have set the occupancies of the Fe found in the active site in some of the monomers of the co- atoms to 50%, which gives much more reasonable B factors. crystals (Fig. 5). The density in the active site was insufﬁcient These results imply that the Fe atom is bound but not tightly to unambiguously place atrazine. For this reason, atrazine coordinated in the active site, consistent with the unusually docking was investigated using extensive molecular-docking high observed Kd of AtzA for iron (5 mM; Noor et al., 2014). calculations with the RosettaLigand docking protocol (Davis Interestingly, Fe2+ bound in related amidohydrolases (e.g. et al., 2009). The results of these calculations suggest that cytosine deaminases; PDB entries 1ra0, 1k70 and 4r88; Ireton nonproductive binding modes are common for atrazine. et al., 2001; Mahan et al., 2004) is also found to be less tightly coordinated than Zn2+ cations, despite Fe2+ being the physiologically relevant cation. There are a number of Ramachandran outliers associated with the active site, His66, His276, Glu298, Asp327 and Asn332, which all have good electron density (Fig. 5) and are unambiguously placed. His66, His276 and Asp327 bind the Fe2+ in the active site directly and introduce strain into the active site. Most of the other Ramachandran outliers are also in good electron density. The one exception is Glu251, which is in a region of weak electron density and is likely to be mobile in the protein. This mobile region sits over the substrate and

Figure 6 Simulated docking of atrazine in the AtzA active site: the highest scoring docked pose produced using the RosettaLigand docking protocol with the Figure 7 coordinates from PDB entry 4v1x that placed atrazine within a physically Differences between AtzA and the melamine deaminase TriA. The reasonable distance for nucleophilic substitution by the activated water. positions of Phe84, Val92, Asp125, Thr217, Thr219, Ile253, Gly255, Although it is notionally possible to rotate the plane of the atrazine ring Asn328 and Ser331, which distinguish AtzA from TriA, are shown in an through 180, the steric environment of the binding pocket appear to AtzA monomer (coloured by secondary structure). The active-site metal prohibit this alternative binding mode. The crystal structure (grey) and (Fe2+) is shown as an orange sphere and the surface of the active site/ the modelled structure produced during docking simulations (cyan) are solvent-access channel is shown. The images are rotated 180 around the shown. The active-site metal (Fe2+) is shown as a blue sphere. vertical axis with respect to each other.

716 Peat et al. AtzA Acta Cryst. (2015). D71, 710–720 research papers

Figure 8 Plausible reaction mechanisms for AtzA. Two plausible reaction mechanisms are proposed involving either bidentate (a) or mondentate (b) coordination of the atrazine Cl atom to the Fe2+ centre of the AtzA active site.

Acta Cryst. (2015). D71, 710–720 Peat et al. AtzA 717 research papers

A high-scoring pose was produced that positioned the residues: Leu84 (Phe), Leu92 (Val), Asp125 (Glu), Ile217 appropriate C atom of atrazine for nucleophilic attack by the (Thr), Pro219 (Thr), Leu253 (Ile), Trp255 (Gly), Asp328 (Asn) metal-bound water. In contrast to our previously proposed and Cys331 (Ser) (AtzA residues shown in parentheses; mechanism based upon homology modelling (Scott et al., Seffernick et al., 2001). Of these nine amino-acid residues, six 2009), the N-ethyl and N-isopropyl moieties of the substrate are part of the substrate-binding pocket/access channel in are oriented away from the metal centre, making hydrophobic AtzA: Phe84, Val92, Thr217, Thr219, Asn328 and Ser331 (Fig. contacts with Val92, Trp87, Leu88, Tyr85 and Phe84 (Fig. 6). 7). Phe84 presents a different rotamer to that found in the The presence of Phe84, Asn328 and Ser331 in the substrate- substrate-free crystal structure. The Cl atom and a ring N atom binding pocket is unsurprising, as these three amino acids had of atrazine are coordinated to the Fe2+, which positions the previously been shown to be the major determinants of atra- substrate ideally for nucleophilic attack by the coordinated zine/melamine specificity in AtzA and TriA, respectively nucleophile. (Noor et al., 2012; Scott et al., 2009; Raillard et al., 2001). In To determine whether the bidentate coordination mode is particular, Ser331 and Asn328 are located within the vicinity likely or whether there is monodentate coordination through of the active-site metal, which is consistent with their the Cl atom, not requiring rotation of His243 but requiring a previously proposed role in catalysis in TriA-mediated slight outwards movement of the hydrophobic pocket melamine deamination and passive promotion of atrazine- described above, would involve further investigation that is mediated atrazine dechlorination (Noor et al., 2012; Scott et beyond the scope of the current work. Regardless, the lateral al., 2009). shift to form the monodentate coordination mode would be In laboratory evolution experiments, altering Thr217 and less than 1 A˚ and would be able to be accommodated without Thr219 has been demonstrated to lead to substantial major reorganization of the active site. improvements in the Km and kcat/Km values of AtzA for There are two hydrogen-bonding interactions from the atrazine (Scott et al., 2009). Thr217 and Thr219 were targeted N-ethyl and N-isopropyl substituents to the side-chain O atom for mutagenesis as homology modelling had predicted their of Asn328 and the side chain of the new rotamer of Glu246, presence in the substrate-binding pocket, and the AtzA respectively. The latter of these residues may play a role in structure presented here confirms this (Fig. 7). Ile253 and stabilizing the negative charge on a ring N atom during the Gly255 are not located in the substrate-binding pocket. formation of the tetrahedral intermediate. The symmetry of Interestingly, however, Ile253 and Gly255 do cluster close to the triazine ring of the subunit could make it possible for the Thr219 (Fig. 7), with Ile253 within 4 A˚ of Thr219. Changes to N-alkyl side chains of atrazine to bind in the opposite orien- the residues at positions 253 and 255 may therefore influence tation to that presented here; however, the additional volume the packing of the residues of the substrate-binding pocket required to accommodate the N-isopropyl side chain (i.e. positions 217 and 219), which is consistent with their compared with the N-ethyl side chain makes this alternate relatively minor role in ‘tuning’ the specificity of the enzyme binding mode unlikely from a steric perspective. for melamine or atrazine (Noor et al., 2012; Raillard et al., The substrate-binding pocket of AtzA appears to be ill-adapted for binding atrazine, with strained conformers of key amino residues, a low-affinity metal-binding site and a high probability of unproductive substrate binding. These observations are consistent with the high Km of wild-type AtzA for atrazine, which exceeds the aqueous solubility of the substrate (i.e. >159 mM; Scott et al., 2009), and may explain why we have been unable to capture clear X-ray structures of the AtzA–atrazine complex.

3.3. Location of the nine amino acids that distinguish AtzA and Figure 9 the melamine deaminase TriA Comparison of AtzA with the alternative atrazine chlorohydrolase TrzN. (a) Superposition of an AtzA The melamine deaminase TriA monomer (green) and a TrzN monomer (cyan). (b) Superposition of the active sites of AtzA (green) and a TrzN monomer (cyan). The active-site metals (Fe2+ and Zn2+; shown in orange and grey, respectively) are differs from AtzA in the iden- shown and the AtzA metal-binding ligands are labelled in green. The unique threonine (Thr325) of TrzN, tities of just nine amino-acid which is equivalent to Asp327 in AtzA, is also labelled in cyan.

718 Peat et al. AtzA Acta Cryst. (2015). D71, 710–720 research papers

2001). The remaining difference between AtzA and TriA is at may do so through combinations of adjusted positioning of the position 125 (Glu in AtzA and Asp in TriA) and has been metal-bound water through shifted hydrogen-bond networks, shown to contribute substantially to the specificity of AtzA altering its pKa, and a changed steric factor in substrate and TriA for dechlorination or deamination (Noor et al., binding and product release. 2012). Although Glu125 is buried (Fig. 7), it does contribute to 3.5. A comparison of the two metal-dependent atrazine the active-site geometry via a hydrogen-bonding interaction chlorohydrolases AtzA and TrzN between the Glu125 side chain and the backbone N atom of Val69, next to His68 which is bound to the Fe2+. The tertiary structure of AtzA is similar to that of TrzN (Seffernick et al., 2010). The monomers of AtzA and TrzN 3.4. The catalytic mechanism of AtzA superpose with a root-mean-square difference of 1.8 A˚ on C We present two alternative potential catalytic mechanisms atoms over 414 residues (Fig. 9a). The metal-binding histidine for the hydrolytic dechlorination of atrazine by AtzA that are side chains and metal overlay (Fig. 9b), with the major constant with the in silico docking presented here, notwith- differences being the longer metal–histidine distances (His66, ˚ ˚ standing that empirical data will be required to test these and His68 and His243 in AtzA at 2.7–3.1 A instead of 2.0–2.1 A in 2+ 2+ previously proposed reaction mechanisms. The first is based TrzN; Fig. 9b), Fe instead of Zn as the metal, the additional 2+ directly upon the interactions shown in the only docking pose His276 interaction with Fe and the additional Fe–Asp327 ˚ that oriented atrazine appropriately for nucleophilic attack interaction (at 2.4–2.5 A). In the TrzN structure, the histidine ˚ and the second is based upon the abovementioned mono- equivalent to His276 (His274) makes a 2.7 A hydrogen bond dentate coordination mode for atrazine (Figs. 8a and 8b, to the water coordinated to the Zn ion. Uniquely among the respectively) that better enables Glu246 to stabilize negative amidohydrolase family, the TrzN structure has a threonine charge on another of the ring N atoms. In both of these residue at position 325, making it the first member of the ninth mechanisms, coordination of the Cl atom to Fe2+ draws elec- subgroup of metal-coordination sites of the amidohydrolase tron density from the C—Cl bond, increasing the susceptibility family (Seibert & Raushel, 2005; Jackson et al., 2014; Khurana of the C atom to nucleophilic attack. In the first mechanism, et al., 2009). the coordinated water is deprotonated by Asp327 followed by Unlike the active site of AtzA, the TrzN active site has nucleophilic attack of the resulting activated hydroxide at the excellent complementarity for its substrate, which almost chlorine-bearing ring C atom. The negative charge of the certainly accounts for the lower Km of TrzN for atrazine tetrahedral intermediate is stabilized on the Fe2+-coordinated (20 mM; Shapir et al., 2002) when compared with that of aromatic N atom, after which chloride is released and AtzA for the same substrate (>159 mM). The mechanism of aromaticity is re-established in the hydroxylated product. TrzN differs from any proposed for AtzA, in which a zinc- The second mechanism with monodentate atrazine coordi- activated water forms the tetrahedral intermediate and highly nation to Fe2+ involves deprotonation of the coordinated activated leaving groups (such as halides) depart readily. water by Glu246, which then rotates to donate a hydrogen Leaving groups with higher pKa values (such as amines) can bond to a ring N atom of atrazine upon substrate binding. This be hydrolyzed by TrzN as they are protonated; Glu241 first N atom then gains the negative charge of the tetrahedral donates a proton to the triazine ring, which then protonates intermediate, stabilized by the hydrogen bond from the the leaving group. Thus, TrzN is an efficient hydrolase with a protonated Glu246, followed by generation of the products as broad substrate range, while AtzA is a less active enzyme that per the first mechanism. can perform either deamination or dehalogenation reactions It is possible that the ‘loose’ coordination geometry of the depending on the identities of the amino-acid residues at Fe2+ has evolved to facilitate the bidentate coordination of positions 328 and 331. AtzA variants that are effectively atrazine, which would not be possible with four ideally co- intermediates between AtzA and TriA possess both of these ordinating residues and a coordinated water. His243 in parti- activities at the same time, but with a lower specificity than cular, at over 3 A˚ from the Fe2+, provides sufficient space for either AtzA or TriA (Noor et al., 2012). bidentate coordination, albeit by adopting a higher energy conformation. We note, however, that such a binding mode is unprecedented, and without a crystal structure of the AtzA– 4. Concluding remarks atrazine complex there is no empirical evidence to support it. For the first time since its discovery 20 years ago, we have While the mechanisms for atrazine dechlorination that we been able to determine the structure of AtzA, the ‘archetypal’ have proposed, in light of the crystal structure and docking, atrazine chlorohydrolase. AtzA appears to be structurally ill- differ from those that we have previously reported based upon adapted to perform its physiological function, with a metal- homology modelling and manual placement of the substrate, coordination geometry that leads to a low affinity for its metal the newly proposed mechanisms remain consistent with the cofactor and a substrate-binding pocket that appears to be empirically determined biochemical and mutagenesis data unable to accept its physiological substrate without substantial reported elsewhere (de Souza et al., 1996; Raillard et al., 2001; reorganization. In addition, four of the active-site amino-acid Scott et al., 2009; Noor et al., 2012). The sequential mutations residues appear to be held in strained conformations, reported earlier (Scott et al., 2009) that reduce the atrazine suggesting that the active site is far from optimized, which is dechlorination activity and increase the deaminase activity consistent with the relatively poor kinetic performance of

Acta Cryst. (2015). D71, 710–720 Peat et al. AtzA 719 research papers

4 1 1 AtzA (kcat/Km of 1 10 M s and a Km in excess of the Mahan, S. D., Ireton, G. C., Knoeber, C., Stoddard, B. L. & Black, substrate’s solubility limit of 153 mM; Scott et al., 2009; Noor et M. E. (2004). Protein Eng. Des. Sel. 17, 625–633. al., 2012). AtzA is certainly less well adapted to its physiolo- Martinez, B., Tomkins, J., Wackett, L. P., Wing, R. & Sadowsky, M. J. (2001). J. Bacteriol. 183, 5684–5697. gical function than the alternative atrazine chlorohydrolase McCoy, A. J., Grosse-Kunstleve, R. W., Adams, P. D., Winn, M. D., 5 1 1 TrzN (kcat/Km of 1 10 M s and a Km value of 20 mM; Storoni, L. C. & Read, R. J. (2007). J. Appl. Cryst. 40, 658–674. Jackson et al., 2014), which has good complementarity for its Mulbry, W. W., Zhu, H., Nour, S. M. & Topp, E. (2002). FEMS substrate and a catalytic mechanism that permits a broader Microbiol. Lett. 206, 75–79. substrate range than AtzA. Both atzA and trzN are prone to Murshudov, G. N., Skuba´k, P., Lebedev, A. A., Pannu, N. S., Steiner, R. A., Nicholls, R. A., Winn, M. D., Long, F. & Vagin, A. A. (2011). horizontal gene transfer, as they are located on transmissible Acta Cryst. D67, 355–367. plasmids (de Souza et al., 1998; Sajjaphan et al., 2004), and it Newman, J., Peat, T. S., Richard, R., Kan, L., Swanson, P. E., will be interesting to see whether the catalytically superior Affholter, J. A., Holmes, I. H., Schindler, J. F., Unkefer, C. J. & TrzN begins to outcompete and replace AtzA in the envir- Terwilliger, T. C. (1999). Biochemistry, 38, 16105–16114. onment. Noor, S., Changey, F., Oakeshott, J. G., Scott, C. & Martin-Laurent, F. (2014). Biodegradation, 25, 21–30. Acknowledgements Noor, S., Taylor, M. C., Russell, R. J., Jermiin, L. S., Jackson, C. J., Oakeshott, J. G. & Scott, C. (2012). PLoS One, 7, e39822. We would like to thank Drs Tara Sutherland and Judith Peat, T. S., Balotra, S., Wilding, M., French, N. G., Briggs, L. J., Scoble (CSIRO) for their constructive comments during the Panjikar, S., Cowieson, N., Newman, J. & Scott, C. (2013). Mol. preparation of this manuscript. We thank the beamline Microbiol. 88, 1149–1163. scientists of the Australian Synchrotron and the CSIRO Raillard, S., Krebber, A., Chen, Y., Ness, J. E., Bermudez, E., Collaborative Crystallization Centre (http://www.csiro.au/C3), Trinidad, R., Fullem, R., Davis, C., Welch, M., Seffernick, J., Wackett, L. P., Stemmer, W. P. C. & Minshull, J. (2001). Chem. Biol. Melbourne, Australia. We also thank OpenEye Scientific 8, 891–898. Software for a license for Afitt, QuacPac and Omega. Russell, R. J., Scott, C., Jackson, C. J., Pandey, R., Pandey, G., Taylor, M. C., Coppin, C. W., Liu, J. W. & Oakeshott, J. G. (2011). Evol. References Appl. 4, 225–248. Sadowsky, M. J., Tong, Z., de Souza, M. & Wackett, L. P. (1998). J. Balotra, S., Newman, J., Cowieson, N., French, N. G., Campbell, P. M., Bacteriol. 180, 152–158. Briggs, L. J., Warden, A. C., Easton, C. J., Peat, T. S. & Scott, C. Sajjaphan, K., Shapir, N., Wackett, L. P., Palmer, M., Blackmon, B., (2015). Appl. Environ. Microbiol. 81, 470–480. Tomkins, J. & Sadowsky, M. J. (2004). Appl. Environ. Microbiol. 70, Balotra, S., Newman, J., French, N. G., Briggs, L. J., Peat, T. S. & Scott, 4402–4407. C. (2014). Acta Cryst. F70, 310–315. Scott, C., Jackson, C. J., Coppin, C. W., Mourant, R. G., Hilton, M. E., Boundy-Mills, K. L., de Souza, M. L., Mandelbaum, R. T., Wackett, Sutherland, T. D., Russell, R. J. & Oakeshott, J. G. (2009). Appl. L. P. & Sadowsky, M. J. (1997). Appl. Environ. Microbiol. 63, Environ. Microbiol. 75, 2184–2191. 916–923. Seffernick, J. L., Aleem, A., Osborne, J. P., Johnson, G., Sadowsky, Cameron, S. M., Durchschein, K., Richman, J. E., Sadowsky, M. J. & M. J. & Wackett, L. P. (2007). J. Bacteriol. 189, 6989–6997. Wackett, L. P. (2011). ACS Catal. 1, 1075–1082. Seffernick, J. L., de Souza, M. L., Sadowsky, M. J. & Wackett, L. P. Copley, S. D. (2000). Trends Biochem. Sci. 25, 261–265. (2001). J. Bacteriol. 183, 2405–2410. Cowtan, K. (2006). Acta Cryst. D62, 1002–1011. Seffernick, J. L., Erickson, J. S., Cameron, S. M., Cho, S., Dodge, A. G., Davis, I. W., Raha, K., Head, M. S. & Baker, D. (2009). Protein Sci. 18, Richman, J. E., Sadowsky, M. J. & Wackett, L. P.(2012). J. Bacteriol. 1998–2002. 194, 4579–4588. Emsley, P., Lohkamp, B., Scott, W. G. & Cowtan, K. (2010). Acta Seffernick, J. L., Reynolds, E., Fedorov, A. A., Fedorov, E., Almo, Cryst. D66, 486–501. S. C., Sadowsky, M. J. & Wackett, L. P. (2010). J. Biol. Chem. 285, Evans, P. R. (2011). Acta Cryst. D67, 282–292. 30606–30614. Frisch, M. J. et al. (2009). Gaussian 09. Gaussian Inc., Wallingford, Seibert, C. M. & Raushel, F. M. (2005). Biochemistry, 44, 6383–6391. Connecticut, USA. Shapir, N., Osborne, J. P., Johnson, G., Sadowsky, M. J. & Wackett, Fruchey, I., Shapir, N., Sadowsky, M. J. & Wackett, L. P. (2003). Appl. L. P. (2002). J. Bacteriol. 184, 5376–5384. Environ. Microbiol. 69, 3653–3657. Shapir, N., Pedersen, C., Gil, O., Strong, L., Seffernick, J., Sadowsky, Hall, R. S., Fedorov, A. A., Marti-Arbona, R., Fedorov, E. V., Kolb, P., M. J. & Wackett, L. P. (2006). J. Bacteriol. 188, 5859–5864. Sauder, J. M., Burley, S. K., Shoichet, B. K., Almo, S. C. & Raushel, Shapir, N., Rosendahl, C., Johnson, G., Andreina, M., Sadowsky, M. J. F. M. (2010). J. Am. Chem. Soc. 132, 1762–1763. & Wackett, L. P. (2005). Appl. Environ. Microbiol. 71, 2214– Ireton, G. C., McDermott, G., Black, M. E. & Stoddard, B. L. (2001). 2220. J. Mol. Biol. 315, 687–697. Shapir, N., Sadowsky, M. J. & Wackett, L. P. (2005). J. Bacteriol. 187, Jackson, C. J., Coppin, C. W., Carr, P. D., Aleksandrov, A., Wilding, 3731–3738. M., Sugrue, E., Ubels, J., Paks, M., Newman, J., Peat, T. S., Russell, Souza, M. L. de, Sadowsky, M. J. & Wackett, L. P. (1996). J. Bacteriol. R. J., Field, M., Weik, M., Oakeshott, J. G. & Scott, C. (2014). Appl. 178, 4894–4900. Environ. Microbiol. 80, 4003–4011. Souza, M. L. de, Wackett, L. P. & Sadowsky, M. J. (1998). Appl. Kabsch, W. (2010). Acta Cryst. D66, 125–132. Environ. Microbiol. 64, 2323–2326. Khurana, J. L., Jackson, C. J., Scott, C., Pandey, G., Horne, I., Russell, Udikovic´-Kolic´, N., Scott, C. & Martin-Laurent, F. (2013). Appl. R. J., Herlt, A., Easton, C. J. & Oakeshott, J. G. (2009). Biochem. J. Microbiol. Biotechnol. 96, 1175–1189. 418, 431–441. Verschueren, K. H. G., Seljeé, F., Rozeboom, H. J., Kalk, K. H. & Kube, M. et al. (2013). Nature Commun. 4, 2156–2159. Dijkstra, B. W. (1993). Nature (London), 363, 693–698. Lennox, E. S. (1955). Virology, 1, 190–206. Wackett, L. P. (2009). Curr. Opin. Microbiol. 12, 244–251.

720 Peat et al. AtzA Acta Cryst. (2015). D71, 710–720